Construction and Building Materials 149 (2017) 741–748

Contents lists available at ScienceDirect

Construction and Building Materials

journal homepage: www.elsevier.com/locate/conbuildmat

Understanding the relationship between the segregation of concrete

and coarse aggregate density and size
Ivan Navarrete a, Mauricio Lopez a,b,⇑
a
Department of Construction Engineering and Management, School of Engineering, Pontificia Universidad Catolica de Chile, Vicuña Mackenna 4860, Casilla 306, Correo 22,
Santiago, Chile
b
Center for Sustainable Urban Development (CEDEUS), Pontificia Universidad Catolica de Chile, Santiago, Chile

h i g h l i g h t s

Segregability of concrete with different CA RSSA and |Dq| are presented.

Concrete segregation is mainly related to the interaction between CA RSSA and |Dq|.

The segregation rate of concrete was shown to be independent of the time of vibration.

A VI of 20% is proposed as a conservative limit for acceptable segregation in concrete.

CA RSSA explained the concrete segregation rate more precisely than MSA.

a r t i c l e i n f o a b s t r a c t

Article history: Segregation of aggregate, which ultimately influences the strength and durability of concrete, is one of
Received 16 December 2016 the major problems during construction. Two factors and their effects on the segregation of fresh con-
Received in revised form 16 May 2017 crete under vibration were studied. Based on the statistical analysis of the experimental results, it was
Accepted 21 May 2017
concluded that the observed rate of segregation is an intrinsic property of concrete and is independent
Available online 31 May 2017
of the vibration time applied. The segregation tendency of a concrete mixture is mainly explained by
the interaction between the specific surface of coarse aggregate and the difference in density between
Keywords:
the aggregate and mortar phase rather than by each individual factor independently.
Fresh concrete
Vibration
Ó 2017 Elsevier Ltd. All rights reserved.
Aggregate size
Aggregate density
Segregation rate
Image analysis
Stereology

1. Introduction mixture is being handled, placed, and transported. On the other

Segregation, the tendency for coarse aggregate to separate from
mortar, remains one of the major problems in fresh concrete. The
consequences of segregation are numerous and may affect the
strength and durability of structures [1]. However, the construc-
tion workforce is mainly unskilled [2]. Thus, control of the mixture
design is necessary to assess a good final quality of the
construction.
ACI 238 defines two kind of segregation, which are dynamic and
static segregation [1]. The dynamic segregation happened when a
⇑ Corresponding author at: Department of Construction Engineering and Man-
agement, School of Engineering, Pontificia Universidad Catolica de Chile, Vicuña
Mackenna 4860, Casilla 306, Correo 22, Santiago, Chile.
E-mail address: mlopez@ing.puc.cl (M. Lopez).
hand, the static segregation happened when a mixture is not
flowing.
The rheology of fresh concrete is complex owing to its multi-
composition and the changes in properties upon hydration. Previ-
ous researchers [3] have shown that the flow behavior of fresh con-
crete can be reasonably approximated by the Bingham model:
s ¼ s0 þ gp  c_ ð1Þ
where s is the shear stress, s0 is the yield stress, gp is the plastic vis-
cosity, and c_ is the shear rate. Therefore, the flow of concrete can be
described by two parameters: yield stress and plastic viscosity.
Fresh concrete may be considered as a two-phase composite
material with coarse aggregate particles in a mortar matrix. The
settlement of a particle in a fluid with Bingham plastic behavior
has been predicted by Beris et al. [4]. They concluded that a sphere

http://dx.doi.org/10.1016/j.conbuildmat.2017.05.185
0950-0618/Ó 2017 Elsevier Ltd. All rights reserved.
742 I. Navarrete, M. Lopez / Construction and Building Materials 149 (2017) 741–748
will settle when the yield stress parameter (Yg), defined in Eq. (2),
is less than 0.143, assuming that the particle density is higher than
that of the fluid.
3  s0
Yg ¼
2  R  jDq j  g
where jDq j is the density difference between the particle and the
fluid, R is the radius of the particle, and g is the gravitational accel-
eration. Thus, in fresh concrete, the beginning of coarse aggregate
settlement is related to the yield stress of the mortar, the density
difference between coarse aggregate and mortar, and the size of
the coarse aggregate. Once the settlement starts, a spherical particle
will sink into the Bingham fluid with a velocity U, which may be
derived from Stoke’s drag equation [5]:
2 R 2  j Dq j  g
U¼ 
9 gp  C s
where gp is the plastic viscosity of the fluid, and Cs is Stoke’s drag
coefficient. Therefore, the velocity of the aggregate settlement is
directly dependent on the difference in density between the coarse
aggregate and the mortar and the size of coarse aggregate, and it is
inversely dependent on the plastic viscosity and drag coefficient.
When the aggregates’ density is lower than that of the mortar, the
principal parameters that affect the stability of the fresh concrete
are the same; i.e., Eq. (3) is still valid and predicts an upward move-
ment of coarse aggregate [6].
Tattersall and Baker [7] showed that when vibration is applied,
there is a significant reduction in the yield stress and a decrease in
the viscosity of the concrete. However, de Larrard et al. [8] found
that the plastic viscosity is unaffected by vibration, which is consis-
tent with previous research [6,9,10] that found that concrete vis-
cosity is more important than yield stress for concrete
segregation during vibration.
Petrou et al. [5] studied the aggregate settlement in concrete in
real time using a scintillation camera to observe and record the set-
tlement of radioactively ‘‘tagged” aggregate in mortar and concrete
during vibration. They found a linear relationship between aggre-
gate settlement and vibration time.
Chia et al. [6] and Petrou et al. [5] studied the settlement of
lightweight and heavyweight aggregates in concrete, respectively.
Both found that concrete mixtures that present a higher difference
between the densities of coarse aggregate and mortar have a
greater segregation tendency.
The relationship between aggregate size and shape and segre-
gation tendency has been studied in different types of concrete
[5,6,11,12]; such as, normal-weight, lightweight and self-
compacting concrete. Shen et al. [11] showed that a reduction of
30% in the maximum size of aggregate (MSA) greatly reduces
dynamic segregation. The lower segregation tendency shown in
mixtures with lower MSA was mainly attributed to the increased
drag force provided by mortar on smaller aggregates, which have
a higher specific surface (i.e., surface area-to-volume ratio). In con-
trast, Esmaeilkhanian et al. [12] compared dynamic segregation in
concrete using either crushed aggregate or rounded aggregate with
similar particle size distribution and found that the segregation
tendency was not significantly different between the two mixtures.
This result was explained by the fact that the aggregate surface-to-
volume ratio does not change considerably between the two
mixtures.
The relationship between the aggregate properties and segrega-
tion tendency of concrete during vibration has been studied by
several researchers over the last decade. However, the combined
effect or interaction between coarse aggregate properties requires
further investigation.
2. Research significance
The stability of fresh concrete without segregation is an impor-
tant issue to be considered for concrete mixture design. The aim of
this research is to assess the combined effects of coarse aggregate
ð2Þ
size, shape and the difference in density between coarse aggregate
and mortar on the dynamic segregation of fresh concrete under
vibration. This will provide a more adequate understanding and
estimate of the intrinsic segregation tendency of a concrete
mixture.
3. Materials and methods
3.1. Material properties and mixture proportion
Ordinary Portland cement (OPC) with a specific gravity of 3.14 and Blaine fine-
ness of 410 m2/kg was used, and a natural river sand with a fineness modulus of
ð3Þ
3.18 was used as fine aggregate (FA) for all mixtures. The absorption of FA was
0.97%, and the specific gravity was 2.72 at the saturated surface dry (SSD) condition.
Additionally, high-range water reducer admixture (HRWA) was used in a dosage of
0.25% by cement weight for all mixtures.
A normal-weight aggregate (NWA)—namely, gravel—and three lightweight
aggregates (LWA)—namely, expanded shale, expanded clay and expanded polystyr-
ene—were used as coarse aggregate (CA). The four types of CAs were sieved to
obtain different single-sized aggregates. The physical properties of single-sized
NWA and LWA used in the study are given in Table 1. The absorption of LWA
was obtained after 72 h immersion to maximize the pore saturation [13].
Stereology, a well-known technique used for interpreting three-dimensional
characteristics of materials based on two-dimensional cross sections, was chosen
to quantify the aggregate specific surface area (SSA) of CA. The estimation of SSA
was made from vertical uniform random (VUR) sections using an unbiased stereol-
ogy technique based on cycloids. Cycloids were used because they are considered to
be isotropic lines on VUR sections in 3D space [14]. SSA is estimated from Eq. (4).
P
I Vc
SSA ¼ 2  P  ð4Þ
P  l Va
P P
where I is the number of intersections, P is the number of points counted, l is the
length of cycloid per point, V c is the volume of concrete of the sample, and V a is the
volume of aggregate of the sample. Since, Vc and Va are given by the mixture design,
P
I
the actual procedure is based on the estimation of P Pl.
To illustrate the estimation of SSA from VUR sections, consider the vertical sec-
tion shown in Fig. 1. A cycloidal test system has been randomly placed on the
image. The test system has a known length of cycloid per point (l). In order to esti-
mate SSA, two counts need to be made on this figure:
a. The number of intersections between the test lines and the boundary of
interest (I). For example, in the case showed in Fig. 1 there are 20
intersections.
b. The number of points that land within the reference space (P). For example,
in the case showed in Fig. 1 there are 16 points.
In this study, two series of concrete mixtures were prepared. Series I consisted
of ten mixtures used to assess the relationship between the segregation of concrete
under vibration and coarse aggregate size and the difference in density between
coarse aggregate and mortar. Series II consisted of two additional mixtures used
to validate the findings and the relationship established with the concretes of Series
I. The concrete mixtures had a water–cement ratio (W/C) of 0.45 and consisted of
approximately 70% mortar and 30% coarse aggregate by volume; the proportion
of each constituent was kept constant in all mixtures. Table 2 presents the mixture
proportions of both series. The moisture states of FA and CA in the given mixture
proportions were SSD.
3.2. Mortar characterization
Table 3 lists the mixture proportions and physical properties of the mortar used
(M1). The rheological behavior of the mortars is determined by the W/C ratio and
the ratio of sand to mortar (s/m) [15]. Previous researchers established procedures
for computing normal-weight [16] and lightweight [17] concrete. According to
these procedures, the ratio of s/m and the ratio of W/C of M1 are similar to those
of conventional mortars.
Several authors [5,9,10,18] have shown that the segregation tendency of con-
crete is related to its mortar viscosity and has no relation to its mortar yield stress.
However, Hafidi et al. [19] established that a direct relation exists between the mor-
tar viscosity and its V-funnel flow time (V-time) [1]. Therefore, the V-funnel test
 I. Navarrete, M. Lopez / Construction and Building Materials 149 (2017) 741–748 743

Table 1
Physical properties of single-sized aggregates.

Aggregate Size (mm) SSA (1/mm) SSD Density (kg/m3) Absorption, 72 h (%)
Gravel 19.0–25.4 0.69 2600 2.44
12.7–19.0 0.97 2630 1.74
9.5–12.7 1.59 2710 1.29
4.8–9.5 1.74 2720 0.99
Expanded Shale 12.7–19.0 0.90 1370 12.47
9.5–12.7 1.47 1410 12.21
4.8–9.5 1.76 1470 12.01
Expanded Clay 9.5–12.7 1.38 950 22.72
4.8–9.5 2.02 980 21.48
2.4–4.8 2.72 1070 20.76
Expanded Polystyrene 4.8–9.5 1.98 15 –
2.4–4.8 2.48 15 –

+ + + +

+ + + +

+ + + +

+ + + +
Intersection (I)
+ Point (P)

Fig. 1. Vertical section with a cycloidal test grid superposed.

Table 2
Concrete mixture proportions.

Mixture ID OPC (kg/m3) Water (kg/m3) FA (kg/m3) CA (kg/m3) HRWA (kg/m3) CA type CA size (mm)
Series I G-1 440 200 950 780 1.1 Gravel 19.0
G-2 440 200 950 789 1.1 Gravel 12.7
ES-1 440 200 950 411 1.1 Exp. Shale 12.7
ES-2 440 200 950 423 1.1 Exp. Shale 9.5
ES-3 440 200 950 441 1.1 Exp. Shale 4.8
EC-1 440 200 950 285 1.1 Exp. Clay 9.5
EC-2 440 200 950 294 1.1 Exp. Clay 4.8
EC-3 440 200 950 321 1.1 Exp. Clay 2.4
EP-1 440 200 950 4.5 1.1 Exp. Pol. 4.8
EP-2 440 200 950 4.5 1.1 Exp. Pol. 2.4
Series II G-3 440 200 950 813 1.1 Gravel 9.5
G-4 440 200 950 816 1.1 Gravel 4.80

Table 3
Mortar mixture proportioning.

Sample ID Water (kg/m3) Cement (kg/m3) Sand (kg/m3) HRWA (kg/m3) W/C s/m Density (Ton/m3)
M1 291 635 1376 1.58 0.46 0.60 2.31

was used to characterize the rheological behavior of M1, which showed a V-time of
6.8 s. This result is similar to the V-times reported by Safawi et al. [9] for mixtures of
conventional mortars.
3.3. Segregation measurement
Mouton [14] showed that the area of an object on random surfaces cut through
the reference space is proportional to the 3D volume of the object in the reference
space. Shen et al. [11] and Kwasny et al. [20] showed that estimating area by pixel
counting is an adequate method to quantify aggregate distribution in concrete sam-
ples. Additionally, Gundersen et al. [21] showed that point counting is an efficient
alternative to estimating area by pixel counting. Therefore, the estimation of aggre-
gate volume fraction at different heights was made with an unbiased stereology
technique based on point counting.
Fifteen 10 cm  20 cm cylinders were cast for each mixture and were vibrated
for different vibration times of 0, 30, 60, 90 and 120 s. The vibration table used had a
fixed frequency of 50 Hz, and the maximum acceleration measured during vibration
was 12 g. At the age of 2 days, the concrete specimens were saw-cut through the
744 I. Navarrete, M. Lopez / Construction and Building Materials 149 (2017) 741–748

+ + + +

+ + + +

+ + + +

+ + + +

+ Point

Fig. 2. An illustration of volume fraction estimation using the point-counting technique.

(a) VI = 20% (b) VI = 17% (c) VI = 15%

Fig. 3. Test specimens with a V.I. lower than 20%

(a) VI = 36% (b) VI = 40% (c) VI = 30%

Fig. 4. Test specimens with a VI higher than 30%

 I. Navarrete, M. Lopez / Construction and Building Materials 149 (2017) 741–748 745

Table 4
Aggregate parameters and stratification coefficients of Series I concrete mixtures.

Sample ID Size (mm) RSSA (mm) jDq j (Ton/m3) VI (%)

0s 30 s 60 s 90 s 120 s
G-1 19.05 4.35 0.42 1 33 55.8 84.9 100.5
G-2 12.7 3.10 0.42 0.5 8.4 27.2 34.6 49.5
ES-1 12.7 3.34 0.89 7 70.4 133.3 154.4 183.1
ES-2 9.53 2.04 0.85 2.1 11 17.3 55.4 85.1
ES-3 6.35 1.71 0.79 1.9 10.3 21.2 32.3 61.7
EC-1 9.53 2.17 1.31 5.3 50.1 96.6 150.4 186.6
EC-2 6.35 1.48 1.28 1.5 14.1 41.4 78.7 103.3
EC-3 3.18 1.10 1.19 2.3 9.4 20.1 25.2 37.9
EP-1 6.35 1.52 2.25 6.5 78.7 155.8 187.8 200
EP-2 3.18 1.21 2.25 4.1 40.3 129.4 152.1 186.6

Pai
Table 5 V ai ¼  100% ð5Þ
Prefi
Segregation rate of concrete mixtures.

Sample ID Ut (%/s) where P ai is the sum of the points reaching the CA in section i, Prefi is the sum of the
points reaching section i, and V ai is the CA volume fraction of section i.
30 s 60 s 90 s 120 s Average An example of the point-counting technique used to estimate the CA volume
G-1 1.1 0.93 0.94 0.84 0.95 fraction is showed in Fig. 2. The number of points reaching the reference space
G-2 0.28 0.46 0.38 0.41 0.38 are counted (i.e. 16). In addition, the number of these points that reach the CA
S-1 2.35 2.22 1.71 1.52 1.95 are counted (i.e. 8). Then, the CA volume fraction is calculated (i.e. 8/16 = 0.5 or
S-2 0.37 0.29 0.61 0.71 0.50 50%).
S-3 0.34 0.35 0.35 0.51 0.39 To evaluate segregation, the volumetric index (V.I.), proposed by Esmaeilkha-
EC-1 1.67 1.61 1.67 1.56 1.63 nian et al. [23], was calculated:
EC-2 0.47 0.69 0.87 0.86 0.72
jV at  V ab j
EC-3 0.31 0.33 0.28 0.32 0.31 V:I:ð%Þ ¼ 2   100% ð6Þ
EP-1 2.62 3.1 2.14 1.66 2.38 V at þ V ab
EP-2 1.34 2.15 1.69 1.55 1.68
where V at and V ab are the aggregate volume fractions of the top and bottom sections,
respectively.

Table 6 4. Results and discussions

First proposed model summary and coefficient results of the segregation rate
regression analysis. 4.1. Limit for acceptable volumetric index
Coefficients Standard Error T Statistic P-value 95% C.I. (Þ
a0 2.01 0.32 6.28 0.00 0.65
Esmaeilkhanian et al. [12] measured the segregation of self-
a1 0.63 0.08 7.56 0.00 0.17 consolidating concrete and established that a V.I. of 25% can be
a2 1.47 0.14 10.79 0.00 0.28 considered as a limit for acceptable segregation resistance. In con-
trast, Kwasny et al. [20] studied the stability of semi-lightweight
R2 = 0.76
concrete and proposed that a concrete can be considered to be seg-
regated when the variation of aggregate content between the bot-
tom and top sections exceeds 20%.
3.5
2.25 Fig. 3 shows the tested specimens of mixtures EC-3 (Fig. 3a), EP-
3.0 2 (Fig. 3b) and G-1 (Fig. 3c), which were vibrated for 30 s and pre-
0.89 sent V.I. values lower than 20%. Fig. 4 shows tested specimens of
2.5
2.25 the same mixtures shown in Fig. 3 that were vibrated for 60 s
2.0 1.31 and present VI values higher than 30%. It can be observed that
Ut (%/s)

specimens with V.I. values lower than 20% did not present segrega-
1.5 tion, and specimens with V.I. values higher than 30% presented
0.42
1.28 segregation. Therefore, a conservative V.I. value of 20% can be con-
1.0
0.85 sidered a limit for acceptable segregation.
1.19 0.79 0.42
0.5 Previous studies [5,10] showed that without vibration, typically
sized aggregate does not settle in mortars with similar V-times to
0.0
0.0 1.0 2.0 3.0 4.0 5.0 M1. Furthermore, settlement stops immediately after vibration is
RSSA (mm) terminated. Table 4 shows that when no vibration is applied, all
the tested mixtures present a V.I. value lower than 7.0%, which is
Fig. 5. Effect of RSSA on segregation rate. (corresponding jDq j values are in ton/m3). similar to the results shown by Chia [6] in non-segregated con-
cretes. Therefore, as expected, no static segregation occurred in
the tested mixtures. Accordingly, it can be stated that the segrega-
longitudinal axis, washed from dust, and air-dried in the laboratory. The dry-cut
tion presented on tested specimens was only due to the segrega-
surfaces were photographed and used to measure the distribution of coarse
aggregate. tion produced by vibration.
The photographs of the tested specimens were divided into three equal sections
(top, middle and bottom). For the top and bottom sections, the volume fraction of
4.2. Segregation rate
CA was estimated using the point-counting technique. For a randomly positioned
point grid the number of points reaching the phase of interest (i.e. CA) divided by
the number reaching the whole section gave an unbiased estimate of volume frac- Aggregate parameters and V.I. average values for each Series I
tion [22], that is: mixture and vibration time are summarized in Table 4. The equiv-
746 I. Navarrete, M. Lopez / Construction and Building Materials 149 (2017) 741–748

Fig. 6. Effect of jDq j on segregation rate. (corresponding RSSA values are in mm).
Fig. 7. Linear regression modelling and experimental results of segregation rate.

Table 7 mortar is constant during the vibration process. Therefore, the seg-
Second proposed model summary and coefficient results of the segregation rate
regression analysis.
regation rates of each mixture, summarized in Table 5, are pre-
ferred instead of V.I. values to analyze the effect of the studied
Coefficients Standard Error T Statistic P-value 95% C.I. () aggregate properties on concrete segregation.
b0 1.37 0.24 5.47 0.00 0.49 The Ut values showed in Table 5 were calculated dividing the VI
b1 0.16 0.09 1.79 0.09 0.19 values and vibration times (t) showed in Table 4 and Eq. (8).
b2 0.30 0.21 1.44 0.16 0.42
b3 0.81 0.12 6.48 0.00 0.25 VI
Ut ¼ ð8Þ
R2=0.89 t

4.3. RSSA value vs. Aggregate size

Table 8
Third proposed model summary and coefficient results of the segregation rate The settling velocity of aggregates during vibration is deter-
regression analysis. mined by the relationship between gravitational force and drag
force [10]. Higher values of gravitational force, which is deter-
Coefficients Standard Error T Statistic P-value 95% C.I. ()
mined by the volume of the aggregate and jDq j value, facilitate
c0 0.96 0.13 7.35 0.00 0.27
aggregate settlement. Conversely, higher values of drag force,
c1 0.96 0.06 16.58 0.00 0.12
which is related to the mortar viscosity and aggregate shape,
R2 = 0.88. reduce aggregate settlement. Leith [24] studied the drag force on
nonspherical objects (i.e., aggregates) in viscoelastic flows (i.e.,
mortar). He established that pressure on the surface of the object
alent specific surface area radius (RSSA ) in Table 4 was calculated by determined one-third of the total drag force and that the other
using the following equation: two-thirds are determined by the friction drag from the tangential
shear stress at the object surface. Therefore, the settling velocity of
3
RSSA ¼ ð7Þ an aggregate is influenced more by its RSSA value than by its size.
SSA Two MLR models were developed for the segregation rate; in
where SSA is the specific surface area of the aggregate. This value the first one, the independent variables were RSSA and jDq j, which
represents the ratio between the volume and the surface area of explained 76% of the variability of the segregation rate. In the sec-
the aggregate used on each mixture. The density difference between ond model, the independent variables were the aggregate size and
aggregate and mortar (jDq j) in Table 4 was calculated as the abso- jDq j, which explained 71% of the variability of the segregation rate.
lute value of the difference between SSD density of the aggregate Therefore, the segregation rate variability is better explained by
and the density of fresh mortar. RSSA than by the aggregate size.
A multiple linear regression (MLR) model that relates V.I. values
with the parameters shown in Table 4 was used to assess the 4.4. Linear regression model for segregation rate
relation between V.I. values and vibration time. The four parame-
ters explained 80% of the variability of the V.I. value, and each MLR models were proposed to correlate the segregation rate
one presented a level of significance below 0.05. Moreover, an with RSSA and jDq j. The first proposed model was similar to the
MLR model that relates segregation rate (U t ), calculated as the ratio model used in Section 4.4, which is represented in Eq. (9).
between V.I. value and vibration time with the parameters shown U t:est ¼ a0 þ a1  RSSA þ a2  jDq j ð9Þ
in Table 4, was used to assess the relation between segregation rate
and vibration time. The four parameters explained 77% of the vari- where U t:est is the estimated segregation rate in %/s, RSSA is in mm,
ability of the segregation rate, and only the vibration time is and jDq j is in ton/m3. Additionally, ai includes the MLR coefficients
insignificant with a P-value of 0.45. Therefore, a linear relation (i = 0,1,2). Because this was a screening model and based on the
exists between vibration time and V.I. values. experimental data a l inear relationship between the aggregate
Furthermore, it can be stated that the segregation rate is an properties (i.e. RSSA and jDq j) and concrete segregation rate was
intrinsic property of the concrete and is independent of the vibra- assumed. In addition, it was assumed that there is no correlation
tion time applied. This is in accordance with the results presented between RSSA and jDq j on the segregation rate of concrete. Both
by Petrou [5], who showed that aggregate settling velocity in assumption were tested using a MLR analysis.
 I. Navarrete, M. Lopez / Construction and Building Materials 149 (2017) 741–748
Table 9
Aggregate parameters and stratification coefficients of Series II concrete mixtures.
Sample ID RSSA (mm) jDq j (Ton/m3)
G-3 2.00 0.42
G-4 1.70 0.42
Table 6 presents the first proposed model summary and coeffi-
cient results. The a0 , a1 and a2 values are shown in Table 6. These
values represent the individual contribution of each parameter on
the segregation rate of concrete for a level of significance of 0.05.
With this model, RSSA and jDq j explained 76% of the variability of
the segregation rate of the concrete.
Fig. 5 shows the relationship between the segregation rate and
the aggregate RSSA value of each concrete mixture. The correspond-
ing jDq j value, in ton/m3, is shown next to each data series. In addi-
tion, straight lines were plotted connecting the results of mixtures
with similar jDq j values, so the trends for each aggregate can be
identified. From Fig. 5, it is observed that the segregation rate
was less affected by RSSA in mixtures with lower jDq j values. The
relationship between the segregation rate and jDq j of each concrete
mixture is shown in Fig. 6. The corresponding RSSA value, in mm, is
shown next to each data series. Also, straight lines were plotted
connecting the results of mixtures with similar RSSA values, so the
trends for each aggregate can be identified. Fig. 6 shows that the
segregation rate is more affected by jDq j in concrete mixtures that
present higher RSSA values. Therefore, it can be concluded that there
is a correlation between RSSA and jDq j on the segregation rate of
concrete. This is supported by the work of Shen et al. [11] and
Esmaeilkhanian et al. [12], which showed that segregation is
directly related to the aggregate mass-to-surface area ratio, which
is related to the multiplication of jDq j and RSSA . Thus, the correlation
between jDq j and RSSA can be represented by the product between
the variables in an improved MLR model, as shown in Eq. (10).
U t:est ¼ b0 þ b1  RSSA þ b2  jDq j þ b3  RSSA  jDq j
The second model explained 89% of the variability of the segre-
gation rate, meaning that the inclusion of the product between RSSA
and jDq j is significant. However, Table 7 shows that RSSA and jDq j
are not significant anymore with P-values of 0.09 and 0.16,
respectively.
Therefore, a third model was proposed, using only the product
between RSSA and jDq j shown in Eq. (11).
U t:est ¼ c0 þ c1  RSSA  jDq j
Table 8 shows that an increase in the RSSA  jDq j coefficient also
increased the segregation rate of concrete for a level of significance
below 0.05. The RSSA  jDq j value explained 88% of the variability of
the segregation rate of concrete, which indicates a good prediction
of the segregation from the linear regression model. However, the
model obtained is valid only for mixtures with similar mortar vis-
cosities and volumes of aggregate used in this study.
4.5. Validation and application of the proposed model
When comparing the model estimates of segregation rate using
Eq. (11) with the experimental results (Fig. 7), it is observed that
the model shows a threshold of RSSA  jDq j under which there is
no segregation. For the concrete tested in this study, that threshold
was 1.0 (Ton  mm=m3 ). The existence of a threshold was also
observed by Beris et al. [4], which predicted that a spherical parti-
cle would settle in a fluid with Bingham plastic behavior only when
the yield stress parameter (Eq. (2)) is below 0.143.
747
RSSA  jDq j(Ton  mm=m3 ) VI (%)
30 s 60 s 90 s 120 s
0.84 0.7 3.6 1.0 4.2
0.71 3.7 0.5 0.8 4.5
Two new concrete mixtures (Series II) that present RSSA  jDq j
lower than 1.00 (Ton  mm=m3 ) were tested to validate the pro-
posed model. Table 9 shows that the V.I. values for the four vibra-
tion times applied were lower than 5% in both tested mixtures,
which means that no segregation occurred in these mixtures, as
predicted by the proposed model.
The current threshold depends on the mortar volume and vis-
cosity; therefore, In order to use it in a more general context, it
is necessary to produce two trial mixtures: one mixture with 20-
mm MSA of normal-weight aggregate (called NW); and other mix-
ture with 4-mm MSA expanded polystyrene (called EP). Each Ut is
measured to determine the new threshold c0 of the mortar using
Eq. (12). These two mixtures are proposed because they have con-
siderably different Ut values, which yields a more reliable estima-
tion of the threshold value.
3:42U tNW  1:84U tEP
c0 ¼ ð12Þ
1:58
where U tNW and U tEP are the measured U t values of the NW and EP
mixtures, respectively.
5. Conclusions
This research aimed to understand the main intrinsic factors
explaining segregation in concrete under vibration by investigating
the combined effect of aggregate size and density difference
between coarse aggregate and mortar. Twelve mixtures with four
different aggregate densities and MSA were investigated. The main
ð10Þ
conclusions follow:
(1) A volumetric index of 20% is proposed as a conservative limit
for acceptable segregation in concrete. The concrete mix-
tures analyzed here presented a threshold below which no
segregation was observed. This means that segregation can
be prevented or reduced by the mixture design of concrete.
(2) The segregation rate produced during vibration was shown
to be independent of the time of vibration, so it is an intrin-
ð11Þ
sic fresh property of the concrete mixture and can be used to
characterize its behavior under vibration.
(3) The volume-to-surface area ratio of coarse aggregate
explained the segregation rate of concrete more precisely
than MSA, meaning that the shape of the aggregate plays a
significant role in segregation. That is, segregation can be
reduced by either reducing MSA or increasing the angularity
of coarse aggregate.
(4) As the density difference between coarse aggregate and
mortar increases, the volume-to-surface area ratio becomes
more significant in explaining the segregation rate. This is
relevant when using lightweight and heavyweight aggre-
gates in normal weight mortars and implies that MSA and
the shape of the coarse aggregate must be especially consid-
ered for minimizing or eliminating segregation.
(5) The segregation tendency of a concrete mixture is more
related to the combined effect of its volume-to-surface ratio
and density difference between coarse aggregate and mortar
than by each individual parameter.
748 I. Navarrete, M. Lopez / Construction and Building Materials 149 (2017) 741–748
(6) A simplified practical methodology is proposed to determine
the segregation threshold of a mixture in order to avoid seg-
regation during vibration.
Acknowledgments
The authors gratefully acknowledge the funding provided by
CONICYT/FONDEF D10I1086 and CONICYT-PCHA/MagisterNacio
nal/2015 - 22150974 and the support provided by CEDEUS, CONI-
CYT/FONDAP 15110020. The authors also recognize Cementos
Melon, DICTUC S.A., SIKA S.A., Mauricio Guerra and Luis Gonzalez
for their contribution to this research project.
References
[1] ACI Committee 238, Report on Measurements of Workability and Rheology of
Fresh Concrete, Farmington Hills, 2008.
[2] F.B. Finger, M.S. González, A.P. Kern, Control of Finished Work – Final Quality
Inspection in a Social Housing Project Control de la obra terminada –
inspección final de calidad en un proyecto de interés social, Rev. Ing.
Construcción 30 (2015) 147–153.
[3] G.H. Tattersall, P.F.G. Banfill, The rheology of fresh, Concrete (1983).
[4] A.N. Beris, J.A. Tsamopoulos, R.C. Armstrong, R.A. Brown, Creeping motion of a
sphere through a Bingham plastic, J. Fluid Mech. 158 (1985) 219–244.
[5] M.F. Petrou, K.A. Harries, F. Gadala-Maria, V.G. Kolli, A unique experimental
method for monitoring aggregate settlement in concrete, Cem. Concr. Res. 30
(2000) 809–816, http://dx.doi.org/10.1016/S0008-8846(00)00223-4.
[6] K. Chia, C. Kho, M. Zhang, Stability of fresh lightweight aggregate concrete
under vibration, ACI Mater. J. 102 (2005) 347–354.
[7] G.H. Tattersall, P.H. Baker, The effect of vibration on the rheological properties
of fresh concrete, Mag. Concr. Res. 40 (1988) 79–89.
[8] F. de Larrard, C. Hu, T. Sedran, J.C. Szitkar, M. Joly, F. Claux, F. Derkx, A new
rheometer for soft-to-fluid fresh concrete, ACI Mater. J. 94 (1997) 234–243,
http://dx.doi.org/10.14359/304.
[9] M.I. Safawi, I. Iwaki, T. Miura, A study on the applicability of vibration in fresh
high fluidity concrete, Cem. Concr. Res. 35 (2005) 1834–1845, http://dx.doi.
org/10.1016/j.cemconres.2004.10.031.
[10] M.F. Petrou, B. Wan, F. Gadala-Maria, V.G. Kolli, K.A. Harries, Influence of
mortar rheology on aggregate settlement, ACI Struct. J. 97 (2000) 479–485.
[11] L. Shen, L. Struble, D. Lange, Modeling dynamic segregation of self-
consolidating concrete, ACI Mater. J. 106 (2009) 375–380, http://dx.doi.org/
10.1016/j.conbuildmat.2014.11.010.
[12] B. Esmaeilkhanian, K.H. Khayat, A. Yahia, D. Feys, Effects of mix design
parameters and rheological properties on dynamic stability of self-
consolidating concrete, Cem. Concr. Compos. 54 (2014) 21–28, http://dx.doi.
org/10.1016/j.cemconcomp.2014.03.001.
[13] ASTM Standard C1761/C1761M, Standard Specification for Lightweight
Aggregate for Internal Curing of Concrete, 2013, http://dx.doi.org/10.1520/
C1761.
[14] P.R. Mouton, Principles and Practices of Unbiased Stereology, The Johns
Hopkins University Press, Baltimore, 2002.
[15] S.B. Singh, P. Munjal, N. Thammishetti, Role of water/cement ratio on strength
development of cement mortar, J. Build. Eng. 4 (2015) 94–100, http://dx.doi.
org/10.1016/j.jobe.2015.09.003.
[16] P.K. Mehta, P.J.M. Monteiro, Concrete: Microestructure, Properties, and
Materials, fourth ed., Mc Grawl Hill Education, New York, 2013.
[17] S. Chandra, L. Berntsson, Lightweight Aggregate Concrete – Science,
Technology and Applications, William Andrew Publishing, New York, 2003.
[18] M.I. Safawi, I. Iwaki, T. Miura, The segregation tendency in the vibration of high
fluidity concrete, Cem. Concr. Res. 34 (2004) 219–226, http://dx.doi.org/
10.1016/S0008-8846(03)00249-7.
[19] A. Hafidi, Correlation between viscosity and V – funnel flow, J. Adv. Technol.
Eng. Res. 3 (2013) 103–108.
[20] J. Kwasny, S.M. Asce, M. Sonebi, S.E. Taylor, Y. Bai, K. Owens, W. Doherty,
Influence of the type of coarse lightweight aggregate on properties of
semilightweight self-consolidating, Concrete (2012) 1474–1483, http://dx.
doi.org/10.1061/(ASCE)MT.1943-5533.0000527.
[21] H.J.G. Gundersen, E.B.V. Jensen, K. Kiêu, J. Nielsen, The efficiency of systematic
sampling in stereology – reconsidered, J. Microsc. 193 (1999) 199–211, http://
dx.doi.org/10.1046/j.1365-2818.1999.00457.x.
[22] C.V. Howard, M.G. Reed, Unbiased Stereology, 2nd ed., Advanced Methods,
New York, 2005.
[23] B. Esmaeilkhanian, D. Feys, K.H. Khayat, A. Yahia, New test method to evaluate
dynamic stability of self-consolidating concrete, ACI Mater. J. 111 (2014),
http://dx.doi.org/10.14359/51686573.
[24] D. Leith, Drag on nonspherical objects, Aerosol Sci. Technol. 6 (1987) 153–161,
http://dx.doi.org/10.1080/02786828708959128.